Research On Laplace Matrix-based Partitioning And Mapping Of Large Ontology

As the development of the Semantic Web, the scale of the ontologies becomes more and more large. It is paid more attention by domestic and foreign researchers about how to implement the large sale ontologies partition efficiently and improve the efficiency of ontology reuse and ontology mapping.First, the thesis introduces the background and the significance of the subject including the summary of ontology and ontology mapping. Then it classifies the research methods and research projects about the block of large scale ontologies, and gives a detail description about the representation and the feature of ontology moduleSecond, we proposes an algorithm of large scale ontology partition which based on Laplace matrix. Our algorithm gives a preprocess to the large scale ontology, and creates an ontology dependency graph, then optimizes the edges and the weights of the graph. Based on the model, It builds a Laplace matrix of the dependency graph, gets the Fiedler value and Fiedler vector and then obtains the modules of the large scale ontology. Through combining the anchor similarity and the theme similarity, the similarity of the modules can be calculated. The partition and mapping of the large scale ontologies also can be achieved.Third, aiming at the limitations of the evaluation methods about the quality of ontology modules, we raise a criteria which is based on the ontology module structure and the modules application to evaluate the quality of ontology modules. Then we compare the Encapsulation, the Independence, the Semantic Richness, Redundancy and the Connectivity of the modules et al. to evaluate the partition methods.Finally, according to the research above, our thesis designs and implements the partition experiment about the large scale ontologies. It gets the experimental data from the Tourism ontology which can be downloaded from http://swoogle.umbc.edu/. The algorithm is executed in the end, and it uses the independence of the modules as the criteria of evaluation. The empirical results indicate our method can achieve ontology partition and enhance the mapping efficiency.